Cultures of Saccharomyces cerevisiae Subjected to Diurnal
Temperature Cycles
Marit Hebly,a,b,cDick de Ridder,b,d* Erik A. F. de Hulster,a,bPilar de la Torre Cortes,a,bJack T. Pronk,a,b,cPascale Daran-Lapujadea,b,c Industrial Microbiology Section, Department of Biotechnology, Delft University of Technology, Delft, The Netherlandsa
; Kluyver Centre for Genomics of Industrial Fermentation, Delft, The Netherlandsb
; Netherlands Consortium for Systems Biology, Amsterdam, The Netherlandsc
; Delft Bioinformatics Lab, Department of Intelligent Systems, Delft University of Technology, Delft, The Netherlandsd
Diurnal temperature cycling is an intrinsic characteristic of many exposed microbial ecosystems. However, its influence on yeast physiology and the yeast transcriptome has not been studied in detail. In this study, 24-h sinusoidal temperature cycles, oscillat-ing between 12°C and 30°C, were imposed on anaerobic, glucose-limited chemostat cultures of Saccharomyces cerevisiae. After
three diurnal temperature cycles (DTC), concentrations of glucose and extracellular metabolites as well as CO2production rates
showed regular, reproducible circadian rhythms. DTC also led to waves of transcriptional activation and repression, which in-volved one-sixth of the yeast genome. A substantial fraction of these DTC-responsive genes appeared to respond primarily to changes in the glucose concentration. Elimination of known glucose-responsive genes revealed an overrepresentation of previ-ously identified temperature-responsive genes as well as genes involved in the cell cycle and de novo purine biosynthesis. In-depth analysis demonstrated that DTC led to a partial synchronization of the cell cycle of the yeast populations in chemostat cultures, which was lost upon release from DTC. Comparison of DTC results with data from steady-state cultures showed that the 24-h DTC was sufficiently slow to allow S. cerevisiae chemostat cultures to acclimate their transcriptome and physiology at the DTC temperature maximum and to approach acclimation at the DTC temperature minimum. Furthermore, this comparison and literature data on growth rate-dependent cell cycle phase distribution indicated that cell cycle synchronization was most
likely an effect of imposed fluctuations of the relative growth rate (/max) rather than a direct effect of temperature.
T
emperature directly influences the kinetics of all biochemical reactions and many cellular processes (1,2). Within their tem-perature range for growth, many microorganisms adapt their metabolic networks and biomass composition to optimize their growth rate and viability in response to changing temperatures. Classical examples of temperature adaptation include modifica-tions of membrane fluidity (3,4) and the expression of heat shock proteins that assist in protein folding at high temperatures (5,6). Laboratory studies on microbial temperature responses are based largely on two experimental systems. Acclimation, i.e., the result of complete physiological adaptation to a given tempera-ture, has been studied both in batch cultures and in chemostats (7–9). In cold shock and heat shock experiments, the responses of microorganisms to sudden upshifts or downshifts in temperature are studied, typically over a time period ranging from a few min-utes to a few hours (3,10–13).In natural environments, microorganisms are subjected to sev-eral types of temperature dynamics. In exposed environments, one of the dominant aspects of temperature dynamics is related to circadian changes, with higher temperatures during the day and lower temperatures during the night. These 24-h temperature cy-cles, which are superimposed on seasonal changes, raise a number of interesting questions related to microbial temperature adapta-tion that have hitherto hardly been addressed in microbial physi-ology. For many microorganisms, the time constants of circadian temperature changes are in the same order of magnitude as their generation times. Moreover, the rate at which microorganisms can adapt their physiology to changes in temperature is itself likely to be strongly temperature dependent, for instance, because rates of transcription and translation decrease strongly at low
temper-atures (2,3). It is therefore unclear whether 24-h cycles enable full temperature acclimation or whether they confront microorgan-isms with a “continuous temperature shock” scenario.
The impact of temperature on physiology has been extensively studied in Saccharomyces cerevisiae, a mesophilic microorganism with an optimum growth temperature of circa 33°C (14,15), par-ticularly because several of its industrial applications, such as brewing, wine making, and frozen-dough applications, involve suboptimal temperatures (16–19). Temperature ranges for growth are strain dependent, but most S. cerevisiae strains are able to grow at temperatures between 4°C and 40°C (15). Natural iso-lates of S. cerevisiae are typically associated with exposed environ-ments, e.g., the bark of oak trees (20,21). To understand the phys-iology of this yeast in its natural environment, it is relevant to study its physiology under circadian temperature fluctuations. Furthermore, such studies may contribute to our understanding
Received 7 March 2014 Accepted 5 May 2014 Published ahead of print 9 May 2014 Editor: A. A. Brakhage
Address correspondence to Pascale Daran-Lapujade, P.A.S.Daran-Lapujade@tudelft.nl.
* Present address: Dick de Ridder, Bioinformatics Group, Wageningen University, Wageningen, The Netherlands.
Supplemental material for this article may be found athttp://dx.doi.org/10.1128 /AEM.00785-14.
Copyright © 2014, American Society for Microbiology. All Rights Reserved.
doi:10.1128/AEM.00785-14
on February 17, 2015 by BIBLIOTHEEK TU DELFT
http://aem.asm.org/
of how evolution under circadian temperature cycles has shaped the physiology and the genome of this important industrial mi-croorganism.
Previous work on temperature responses of S. cerevisiae in-cluded many studies on heat and cold shock responses (11,22–28) and on batch cultures grown at different temperatures (29) and analyses with steady-state chemostat cultures (14,30,31). In the latter studies, detailed physiological analysis of the cultures was combined with genome-wide expression analysis, thereby provid-ing inclusive descriptions of fully temperature-acclimated S.
cerevisiae cultures.
The aim of this study is to investigate the impact of diurnal temperature cycles (DTC) on S. cerevisiae and to assess the extent to which these responses can be predicted from steady-state analyses. To this end, we used a continuous-cultivation setup, in which yeast was grown under controlled conditions and subjected to 24-h sinusoidal temperature cycles. This system was recently used to specifically investigate the impact of temperature dynam-ics on yeast glycolysis, based on integrated modeling and experi-mental analysis of the in vivo kinetics of glycolytic enzymes (32). Rather than focusing on a single metabolic pathway, the present study investigates the overall physiology and transcriptome of S.
cerevisiae during circadian temperature cycles fluctuating between
12°C and 30°C. Data from cultures grown under this dynamic temperature regime are compared with data from fully acclimated chemostat cultures.
MATERIALS AND METHODS
Strain and media. The prototrophic haploid yeast strain Saccharomyces cerevisiae CEN.PK113-7D (MATa) (33,34) was used in this study. The medium used for chemostat and sequential batch cultivation contained 0.3 g · liter⫺1(NH4)2SO4, 0.3 g · liter⫺1KH2PO4, 0.5 g · liter⫺1MgSO4·
7H2O, 3.0 g · liter⫺1NH4H2PO4, and 25 g · liter⫺1glucose. The anaerobic
growth factors Tween 80 and ergosterol (0.42 g · liter⫺1and 10 mg · liter⫺1, respectively), trace elements, and vitamin solution (both 1 ml · liter⫺1) were added to the medium as described previously (35). The medium was supplemented with 0.15 g · liter⫺1silicone antifoam (anti-foam C; Sigma-Aldrich, St. Louis, MO, USA).
Chemostat cultivation. Anaerobic glucose-limited chemostat cul-tures were grown in 2-liter laboratory bioreactors (Applikon, Schiedam, The Netherlands) with a working volume of 1.4 liters and at a dilution rate of 0.03 h⫺1. The pH was controlled at 5.0 by the automatic addition of 2 M KOH, and the stirrer speed was kept at 800 rpm. To ensure anaerobic conditions, bioreactors and medium vessels were continuously sparged with pure nitrogen (N2) gas at flow rates of 0.7 liters · min⫺1and circa 7
ml · min⫺1, respectively, and equipped with Norprene tubing (Saint-Gobain Performance Plastics, Courbevoie, France) and Viton O-rings (Eriks, Alkmaar, The Netherlands). Active temperature regulation was performed and measured online by connecting a platinum resistance thermometer, placed in a socket in the bioreactor, to an RE304 low-tem-perature thermostat (Lauda, Lauda-Königshofen, Germany). Steady-state sampling of cultures grown at a constant temperature of 12°C or 30°C was performed when biomass dry weight, metabolite levels, and carbon dioxide production rates differed by⬍5% for at least three con-secutive volume changes.
Cultivation in sequential batch reactors. The use of sequential batch reactors (SBRs), instead of single batches inoculated from standardized shake-flask-grown precultures, has been shown to enhance the reproduc-ibility of specific growth rate measurements (29,36). SBR experiments at 12°C and 30°C were performed under conditions identical to those for chemostat cultivation, with a working volume of 1 liter. The first batch was inoculated with a shake flask preculture grown overnight (37). Growth was constantly monitored via online CO2measurements in the
off-gas with an NGA 2000 analyzer (Rosemount Analytical, Orrville, OH, USA). When the CO2percentage in the off-gas dropped below 0.05%,
indicating depletion of glucose, a computer-controlled peristaltic pump removed 0.9 liters of culture, leaving approximately 0.10 liters as the in-oculum for the next batch (36). Specific growth rates at 12°C and 30°C were calculated based on biomass dry weight measurements from samples taken during the fourth and fifth cycles.
Diurnal temperature experiments. After three volume changes at a constant temperature of 30°C without significant changes in biomass dry weight, a preprogrammed sinusoidal temperature profile, T (°C)⫽ 21 ⫹ 9 sin{[t (h)⫹ 6] /12}, was started. This profile was designed to mimic a real-life circadian temperature cycle recorded in nature (Fig. 1). Samples were taken during the fifth and/or sixth temperature cycle. To minimize disturbance, sampling volumes did not exceed 5% of the reactor volume during a single temperature cycle, and intervals of at least 3 h were main-tained between sampling points.
Analytical methods. Gas analysis and analyses of biomass dry weight, total cell concentration, cell size, and residual glucose and extracellular and intracellular metabolite concentrations in chemostat cultures were performed as described previously (32,37). Samples for the analysis of the intracellular storage carbohydrates trehalose and glycogen were taken in duplicate from two independent replicate cultures. By means of a rapid sampling setup (32), circa 2.2 ml of broth was rapidly quenched in 5 ml of cold (⫺40°C) pure methanol. The exact sampling weight was measured, and samples were washed once with 5 ml of cold (⫺40°C) pure methanol. The pellet was stored at⫺80°C until analysis of glycogen and trehalose concentrations according to a previously reported procedure (38). Glu-cose concentrations were determined with an EnzyplusD-glucose kit
(EZS781⫹; BioControl Systems, Bellevue, WA, USA).
Prediction of the biomass specific glucose consumption rate during DTC. The biomass specific glucose consumption rate, qS(mmol glucose ·
g [dry weight]⫺1· h⫺1), was calculated by using equations 1 and 2. The amount of biomass, MX(grams), in the reactor did not change
substan-tially during DTC; therefore, a constant value of 1.95 g · liter⫺1was used (seeFig. 3C). Changes in the amount of substrate in the culture broth, dMS
dt (mmol · h
⫺1), were calculated based on changes in residual glucose
concentrations (CS) per 3-h interval. Subsequently, the rate of glucose
consumption, rateS, was calculated with equation 2, in which CS,inis the
substrate concentration in the feed (mmol · liter⫺1) and⌽ is the medium flow rate (liters · h⫺1).
FIG 1 Diurnal temperature profile. The temperature profile imposed on che-mostat cultures of S. cerevisiae was derived from a Web-accessible air temper-ature database (http://www.infoclimat.fr/recherche/sphinx.php?q⫽05%2F08 %2F2007). The open symbols are the temperatures measured in Strasbourg, France, on 5 August 2007, and the solid line indicates the programmed sinu-soidal function, T (°C)⫽ 21 ⫹ 9 sin{[t (h) ⫹ 6] /12}.
Hebly et al.
on February 17, 2015 by BIBLIOTHEEK TU DELFT
http://aem.asm.org/
qS⫽ rateS ⁄ MX (1) dMs
dt ⫽ rateS⫹ ⌽
共
CS,in⫺ CS兲
(2) Budding index. For determination of the budding index (BI), i.e., the percentage of cells carrying a bud, multiple microscopic slides were pre-pared from each independent sample, and multiple pictures of each slide were taken by phase-contrast microscopy with an Imager-D1 microscope equipped with an AxioCam MR camera (Carl Zeiss, Oberkochen, Ger-many), using an EC Plan-Neofluar 40⫻/0.75 Ph 2 M27 objective (Carl Zeiss, Oberkochen, Germany). To ensure unbiased analysis, photographs were blindly labeled and randomly shuffled before counting. At least 200 cells were counted for each sample point and used to determine the bud-ding index.Flow cytometry analysis. Cell cycle phase distribution in yeast popu-lations was estimated with a flow cytometry-based method. A sample of culture broth equivalent to circa 1.6⫻ 106cells was taken from the
bio-reactor and centrifuged (5 min at 4,700⫻ g). The pellet was washed once with cold phosphate buffer (3.3 mM NaH2PO4, 6.7 mM Na2HPO4, 130
mM NaCl, 0.2 mM EDTA) (39), vortexed briefly, centrifuged again (5 min at 4,700⫻ g), and suspended in 800 l 70% ethanol while vortexing. After the addition of another 800l 70% ethanol, fixed cells were stored at 4°C until further staining and analysis. Staining of the cells with Sytox green nucleic acid stain (catalog number S7020; Invitrogen) was per-formed as described previously (40). Samples were analyzed on a FACSCalibur instrument (BD Biosciences, Franklin Lakes, NJ, USA) equipped with an argon laser at 488 nm. The cell cycle phase distribution for each sample was estimated by using the histogram modeler tool from the free software CyFlogic (version 1.2.1; CyFlo Ltd., Turku, Finland).
Microarray and transcriptome analyses. Samples for microarray analysis were taken during the fifth and sixth temperature cycles from two independent duplicate cultures. To minimize disturbance, the sampling volume was kept within 5% of the reactor volume during 24 h. For the last time point, one additional analytical duplicate sample was taken, resulting in a complete data set of 13 arrays. The sampling procedure (quenching in liquid nitrogen), processing of the samples, RNA isolation, and microar-ray analysis with the use of the GeneChip Yeast Genome S98 armicroar-ray (Af-fymetrix, Santa Clara, CA, USA) were performed as described previously (41). Microarray data acquisition (using a target value for global scaling of 240), elimination of insignificant variation, and extraction of the 6,383 yeast open reading frames (42) were done as described previously (43). Sample points from the fifth and sixth temperature cycles were combined, resulting in six sample points covering one temperature cycle (at temper-atures of 30°C, 21°C, 14.6°C, 12°C, 21°C, and 27.4°C). The expression data from these 6 time points were subjected to time course differential expres-sion analysis by empirical analysis of digital gene expresexpres-sion data in R (EDGE, v1.1.291) (43,44). All genes with a P value of⬍0.002 were con-sidered to be significantly changed (1,102 genes) and were subsequently k-means clustered using the k-means clustering algorithm of Genedata Expressionist Pro (v3.1) that uses positive correlation as distance metric (43), resulting in six clusters. The clusters were examined by the use of a hypergeometric distribution test (45) for enrichments of functional cate-gories and transcription factor (TF) binding (46), as described previously (43,47). If only a secondary or tertiary category, as defined by the Munich Information Center for Protein Sequences (MIPS) database, was en-riched, this category was shown, and Gene Ontology (GO) leaf categories were shown if they identified a category that was not enriched in MIPS. DTC-specific genes were examined by a hypergeometric distribution test for enrichments of a number of custom-made categories consisting of cell cycle phase marker genes (according to references48and49).
To distinguish between glucose responses and temperature responses, the transcriptome data set was compared with data from previous studies on glucose-dependent transcriptional responses of S. cerevisiae CEN. PK113-7D (50,51). Primary data (.CEL files) from those previous studies were retrieved from the Gene Expression Omnibus database (accession
numbersGSE3821andGSE8187, respectively [http://www.ncbi.nlm.nih .gov/geo]) and uploaded into Genedata Expressionist Pro (v3.1). k-means clustering of the 215 and 195 genes in the two data sets was performed separately as described above (see Fig. S1 in the supplemental material).
Steady-state chemostat cultures at constant temperatures of 12°C and 30°C (independent duplicate cultures at both temperatures) were sam-pled for microarray analysis as described above, thereby expanding the total array set to 17 arrays. Genes that were differentially expressed be-tween two conditions (e.g., pairwise comparison bebe-tween cultures at steady state at 12°C and those at steady state at 30°C) were identified by first removing probe intensity background (52) and quantile normalizing intensity levels (53) as in robust multichip analysis (RMA). Subsequently, array groups (two biologically independent arrays per group) that corre-sponded to the sampling point were compared based on the perfect-match probe intensity levels only (52), by performing a per-probe-set two-way analysis of variance (with the factors “probe” and “sampling point”). Adjustment of P values for multiple testing was done with Šidák step-down adjustment (54). Differences with adjusted P values of⬍0.05 and a fold difference of 2 or higher were considered significant.
Principal component analysis. For each sampling point during DTC, an average expression profile was calculated based on the two biologically independent arrays. This set was extended with average expression pro-files of steady-state samples at 12°C and 30°C. Principal component anal-ysis (55) was performed, and samples were projected onto the first two components.
Microarray data accession number. The complete data set (17 arrays) was deposited at the Gene Expression Omnibus database (http://www .ncbi.nlm.nih.gov/geo) under accession numberGSE55372.
RESULTS
Rhythmic, reproducible changes in glucose fermentation dur-ing diurnal temperature cycles. To study diurnal temperature cycles (DTC) under controlled laboratory conditions, sinusoidal temperature variations with a 24-h period were imposed on an-aerobic, glucose-limited chemostat cultures of S. cerevisiae CEN.PK113-7D. The temperature profile was designed to closely resemble a natural 24-h temperature cycle ranging between 30°C and 12°C (Fig. 1). The maximum specific growth rates (max) of S. cerevisiae CEN.PK113-7D at 30°C and 12°C were estimated in
sequential batch reactor setups (see Materials and Methods) un-der conditions identical to those of the chemostat cultures but in the presence of excess glucose. At 30°C, themaxwas 0.310⫾
0.002 h⫺1, while at 12°C, it was only 0.048⫾ 0.001 h⫺1. To prevent washout of the chemostat cultures during the temperature mini-mum of the DTC, the dilution rate in the DTC chemostat cultures was set at 0.03 h⫺1.
During a series of consecutive DTC, the biomass concentration in the chemostat cultures remained essentially constant (Fig. 2). However, CO2production revealed a clear cyclic variation in
fer-mentative activity (Fig. 2). As the temperature decreased from 30°C to 12°C during the first half of the 24-h cycles, the CO2
production rate decreased by 35%. The subsequent temperature increase was accompanied by an acceleration of the CO2
produc-tion rate, after which it reached its initial level. The residual glu-cose concentration also markedly fluctuated during the DTC (Fig. 2). For both CO2and residual glucose, the amplitude of this
fluc-tuation decreased as yeast cells acclimated to the DTC and became steady and reproducible after three cycles (Fig. 2).
A closer inspection of the fifth and sixth temperature cycles revealed that the residual glucose concentration was inversely cor-related with temperature (Fig. 3A), reaching values of 2.64⫾ 0.07 mM and 0.20⫾ 0.01 mM at 12°C and 30°C, respectively.
on February 17, 2015 by BIBLIOTHEEK TU DELFT
http://aem.asm.org/
ingly, the glucose concentration profiles were not as symmetrical as the imposed temperature profile. Instead, glucose concentra-tions decreased much faster when the temperature increased after passing the temperature minimum than they increased as the tem-perature minimum was approached. This asymmetry was also vis-ible in the off-gas CO2profile (Fig. 3B), in which a slow
decelera-tion of CO2 production during the temperature decrease was
followed by a sharp acceleration when the temperature increased again. After “overshooting,” the CO2profile resumed its initial
level before the temperature maximum was reached.
Biomass concentration, measured as culture dry weight, re-mained constant at a value of 1.95 g · liter⫺1, with a variation of ⬍5% during the DTC (Fig. 3C). The concentrations of extracel-lular metabolites (ethanol, glycerol, lactate, succinate, and pyru-vate) (Fig. 3DtoG) were largely unaffected by the cyclic temper-ature variation, with the notable exception of acetate, the concentration of which rhythmically varied by circa 70% (Fig. 3H). However, the concentration of acetate was very low com-pared to those of ethanol and glycerol and represented⬍0.24% of the organic metabolites detected in culture supernatants.
Substantial transcriptional reprogramming during DTC. To obtain an overview of the cellular responses of S. cerevisiae to diurnal temperature cycles, a transcriptome analysis was
per-formed, covering six time points during the 24-h temperature cycle (Fig. 4). In contrast to the relatively mild physiological im-pact of the DTC, microarray analysis revealed major reprogram-ming of gene expression. Using stringent statistical criteria (P value of⬍0.002), a set of 1,102 genes (representing 17% of the S.
cerevisiae genome) showed changes in transcript levels. Clustering
of these genes according to their transcript profiles revealed pat-terns of transcriptional induction and repression that either pos-itively or negatively correlated with the temperature profiles (Fig. 4). Transcript levels of 498 genes peaked at the lowest temperature point in the cycle (clusters 1, 2, and 3) (Fig. 4), while transcript levels of 384 genes (clusters 4 and 5) (Fig. 4) decreased with de-creasing temperature and reached their lowest transcript level at the temperature minimum. The expression levels of 220 genes in cluster 6 were hardly affected when the temperature decreased from 30°C to 12°C but were strongly increased as the temperature increased again, reaching the highest level before the temperature maximum was reached, after which expression levels returned to their values at 30°C.
While transcript levels of genes in clusters 1 and 2 were in-versely correlated with temperature, they could be differentiated by the rates at which their transcript levels changed. Transcript levels of genes in cluster 1 increased faster with decreasing tem-FIG 2 Adaptation of S. cerevisiae cultivated in anaerobic glucose-limited cultures to diurnal temperature profiles. (Top) Applied temperature cycles over time (dashed line) and average biomass concentrations (Œ) with standard deviations measured at the highest (30°C) and lowest (12°C) temperatures during seven consecutive cycles. (Bottom) Percentage of CO2in the off-gas measured continuously (with averages represented by the black line and averages⫾ standard
deviations depicted by the gray lines). The dashed line indicates the steady-state CO2percentage at 30°C. The dots represent the average residual glucose
concentrations with standard deviations that were measured at the highest (30°C) and lowest (12°C) temperatures during seven consecutive temperature cycles. For all data, the averages and standard deviations for three independent replicates are represented.
Hebly et al.
on February 17, 2015 by BIBLIOTHEEK TU DELFT
http://aem.asm.org/
peratures between 21°C and 12°C than did those in cluster 2, while the expression levels of genes in cluster 2 displayed a steeper de-crease with increasing temperatures between 12°C and 21°C. Al-though transcript profiles of genes in cluster 3 resembled those of genes in cluster 2, cluster 3 was characterized by a steeper decrease of transcript levels as the temperature increased from 12°C to
21°C. This decrease was followed by an increase in expression levels between 21°C and 30°C, returning to the initial expression values. These differences in transcriptional regulation between clusters 1, 2, and 3 were mirrored in the diversity of the overrep-resented functional categories in these three clusters (Table 1). Cluster 1 (54 genes) was specifically enriched for genes involved in FIG 3 Physiological characterization of S. cerevisiae cultivated under anaerobic glucose-limited conditions during a diurnal temperature cycle. In all panels, the dashed line illustrates the applied temperature profile over time. (A) Residual glucose concentration () and biomass specific glucose uptake rate (qS) (dotted line). (B) CO2percentage in the off-gas (the solid black line represents the average of 13 independent cultures, and the gray solid lines depicts the averages⫾
standard deviations). (C) Culture dry weight. (D) Ethanol concentration. (E) Glycerol concentration. (F) Lactate concentration (o) and succinate concentration (〫). (G) Pyruvate concentration. (H) Acetate concentration. Data points represent the average concentrations and error bars represent the standard deviations of at least seven independent culture replicates.
FIG 4 Transcriptional reprogramming of S. cerevisiae grown under anaerobic glucose-limited conditions during a diurnal temperature cycle. Based on their different time-dependent transcript profiles, the 1,102 genes that were considered to be significantly changed after time course analysis (P value of⬍0.002) (see Materials and Methods) were assigned to six clusters. The expression level of each gene was mean normalized and subsequently averaged over the two independent culture replicates. The dots represent the averages of these averaged expression values for all genes in the respective cluster. The gray lines represent the averages⫾ standard deviations. The dashed line illustrates the applied temperature profile over time.
on February 17, 2015 by BIBLIOTHEEK TU DELFT
http://aem.asm.org/
TABLE 1 Functional enrichment analysisa
Cluster (no. of genes) and
profile Enriched functional category (MIPS functional category)
No. of significantly changed genes in enriched functional category/total
no. of genes in functional category P value
1 (54) Cell wall (42.01) 8/215 4.0⫻ 10⫺4 2 (146) Mbp1 14/165 2.5⫻ 10⫺5 Ino4 6/33 8.6⫻ 10⫺5 Swi6 12/160 2.7⫻ 10⫺4 Swi4 11/144 4.2⫻ 10⫺4 Ino2 5/31 6.1⫻ 10⫺4 3 (298) Protein synthesis (12) 108/511 3.8⫻ 10⫺46 Transcription (11) 113/1,036 1.5⫻ 10⫺20
Protein with binding function or cofactor requirement (16) 81/1,049 1.1⫻ 10⫺6
Fhl1 48/208 2.0⫻ 10⫺21 Rap1 27/145 4.0⫻ 10⫺10 Sfp1 14/50 3.5⫻ 10⫺8 4 (51) Bas1 4/39 2.4⫻ 10⫺4 Gcn4 7/182 5.5⫻ 10⫺4 5 (333) Energy (2) 80/360 4.1⫻ 10⫺31 Mitochondrion (42.16) 31/170 5.5⫻ 10⫺10
C compound and carbohydrate metabolism (1.05) 53/510 6.0⫻ 10⫺7 Transported compounds (substrates) (20.01) 56/585 3.9⫻ 10⫺6
Fe/S binding (16.21.08) 4/5 3.5⫻ 10⫺5
Protein folding and stabilization (14.01) 15/95 1.0⫻ 10⫺4 Mitochondrial transport (20.09.04) 15/100 1.9⫻ 10⫺4 Hap4 19/57 2.7⫻ 10⫺11 Hsf1 15/55 7.5⫻ 10⫺8 Hap3 9/26 3.6⫻ 10⫺6 Cin5 24/180 1.8⫻ 10⫺5 Ume6 17/131 4.2⫻ 10⫺4 6 (220) Metabolism (1) 87/1,530 1.3⫻ 10⫺7
Cell rescue, defense, and virulence (32) 41/558 2.0⫻ 10⫺6 Lysosomal and vacuolar protein degradation (13.13.04) 7/27 2.6⫻ 10⫺5
Energy (2) 27/360 9.7⫻ 10⫺5
Msn2 15/122 1.7⫻ 10⫺5
Msn4 14/121 6.3⫻ 10⫺5
aEach cluster was tested for overrepresentation of genes belonging to functional categories (see Data Set S1 in the supplemental material). Enrichment for MIPS categories is
indicated in lightface type, and enrichment for targets of transcription factors is indicated in boldface type. P values indicate the probability of finding an enrichment by chance.
on February 17, 2015 by BIBLIOTHEEK TU DELFT
http://aem.asm.org/
cell wall organization (8 genes [P value of 4.0⫻ 10⫺4]). Cluster 2 (146 genes) was enriched for targets of the transcriptional activa-tor Ino2/Ino4 (P values of 6.1⫻ 10⫺4and 8.6⫻ 10⫺5, respec-tively), involved in the induction of phospholipid biosynthesis (56). Cluster 2 was enriched for targets of the transcription factors Swi4, Swi6, and Mbp1 (11, 12, and 14 genes [P values of 5.1⫻ 10⫺4, 3.3⫻ 10⫺4, and 2.7⫻ 10⫺5], respectively), involved in cell cycle progression. Cluster 3 (298 genes) showed a very strong overrepresentation of genes involved in protein synthesis, more specifically in ribosome biogenesis and rRNA processing (93 genes [P value of 4.3⫻ 10⫺49] and 73 genes [P value of 2.8⫻ 10⫺53], respectively). Consistent with these observations, cluster 3 also contained a large number of ribosomal protein genes, whose tran-scription is regulated by the trantran-scription factors Fhl1, Rap1, and Sfp1.
Genes in clusters 4 and 5, which were characterized by a down-regulation of expression at low temperatures, differed mostly in their response to increasing temperatures (Fig. 4). While the pression of genes in cluster 4 gradually resumed toward their ex-pression levels at 30°C, genes in cluster 5 exhibited a slight over-shooting of their transcript levels before resuming their initial expression level at 30°C. As observed for clusters 2 and 3, this subtle difference in temperature responses between clusters 4 and 5 coincided with substantial differences in the overrepresentation of functional categories. Cluster 5 was strongly enriched for genes involved in energy metabolism (80 genes [P value of 4.1⫻ 10⫺31]) and more specifically in energy conservation via respiration and energy reserves (38 genes [P value of 4.2⫻ 10⫺19] and 12 genes [P value of 1.3⫻ 10⫺5], respectively). Conversely, few categories were enriched among the 51 genes in cluster 4, covering targets of Bas1 (involved in one-carbon metabolism) (P value of 2.4⫻ 10⫺4) and an enrichment for targets of the Gcn4 transcription factor (7 genes [P value of 5.5⫻ 10⫺4]) involved in the regulation of amino acid biosynthesis. In cluster 6 (220 genes), various cellular func-tions and cell protection mechanisms were overrepresented, in-cluding the subcategories stress response (37 genes [P value of 5.3 ⫻ 10⫺7]), lysosomal and vacuolar protein degradation (7 genes [P value of 2.6⫻ 10⫺5]), and metabolism of energy reserves (9 genes [P value of 6.9⫻ 10⫺5]).
Predominant response to cyclic variations in residual glu-cose concentrations. Cyclic variation of residual gluglu-cose concen-trations was one of the most pronounced phenomena observed when yeast cultures were exposed to DTC. The presence of glu-cose, the preferred carbon source of S. cerevisiae, in culture media is well known to trigger a wide array of transcriptional responses in a concentration-dependent manner (57). Previous studies in-vestigated the transcriptional responses of S. cerevisiae supplied with low and high concentrations of glucose. For instance, Kresnowati et al. and Van den Brink et al., using the same yeast strain as the one used in the present study, identified the response of aerobic glucose-limited cultures of S. cerevisiae to a sudden increase of the glucose concentration (50,51). The sets of genes found to be differentially expressed in these two studies largely overlap and together cover the transcriptional response to glucose excess, such as increased expression levels of genes related to ribo-some biogenesis and rRNA processing and repression of genes involved in energy metabolism and (storage) carbohydrate me-tabolism (50,51). These responses were also identified in the pres-ent work.
A close comparison of the DTC data set and the two previously reported data sets revealed that among the 498 genes whose tran-scription was upregulated during the phases of the DTC when the residual glucose concentration was high (clusters 1, 2, and 3), over half (283 genes) (Fig. 5A) were also upregulated in the studies by Kresnowati et al. and Van den Brink et al. Similarly, about two-thirds of the 604 genes downregulated at high residual glucose concentrations in the DTC (clusters 4, 5, and 6) were also identi-fied as being glucose repressed in the two previously reported data sets (409 genes) (Fig. 5B). These 692 “glucose-responsive” genes were not uniformly distributed over the six clusters (Fig. 5C) and belonged to functional categories previously associated with re-sponses to glucose excess.
To separate indirect transcriptome responses to changes in the glucose concentration from temperature-specific responses, the 692 glucose-responsive genes were removed from the set of DTC-responsive genes. This resulted in a smaller set of 410 genes (Fig. 5AtoC). This set of 410 genes was, however, still strongly enriched for categories that are associated with changes in glucose concentrations (e.g., protein synthesis) (Fig. 5D; see also Data Set S2 in the supplemental material). To further narrow down the data set and more precisely define the set of DTC-responsive genes, the data sets of Kresnowati et al. and Van den Brink et al. were further mined. As those two studies employed rather strin-gent statistical criteria, their complete data sets were used. The profiles of the 410 genes identified in the present study were com-pared with the profiles of the corresponding genes in the studies by Kresnowati et al. and Van den Brink et al. (see Materials and Methods; see also Fig. S1 in the supplemental material). Profiles among these 410 genes that responded to glucose in these pub-lished data sets were removed from the DTC data set (see Fig. S2 in the supplemental material), resulting in a final data set of 253 genes that were considered to be DTC specific. As anticipated, this additional data processing successfully filtered out genes involved in protein synthesis (see Table S1 in the supplemental material). The DTC-specific genes belonged to six main cellular processes and were spread across the different clusters (Table 2; see also Data Set S3 in the supplemental material): lipid metabolism (more spe-cifically phospholipid metabolism), endoplasmic reticulum (ER)-to-Golgi transport, RNA polymerase III transcription, one-car-bon metabolic processes, amino acid metabolism, and cell cycle progression.
Cyclic variations in cell cycle distribution among the yeast population. Among the genes that responded to DTC in an apparently glucose-independent manner were several targets of the transcription factors Swi4, Swi6, and Mbp1 (Table 2). These three transcription factors form complexes (Swi4p-Swi6p forms the SBF complex and (Swi4p-Swi6p-Mbp1p forms the MBF complex) that play an active role in cell cycle control by regulating transcription during the transition from G1 to S
phase of the cell cycle (58). Measurement of cell number and size during the course of a temperature cycle revealed a 22% decrease in cell number and a concomitant 16% increase in cell size during half of the cycle when the temperature decreased, followed by an increase in cell number and a drop in cell size 15 h after the temperature maximum, when the temperature in-creased to a value above 15°C (Fig. 6A). These data suggested that the yeast population, or rather a fraction of the popula-tion, was arresting cell division as the temperature decreased while still growing in size. Furthermore, the data indicated that
on February 17, 2015 by BIBLIOTHEEK TU DELFT
http://aem.asm.org/
cell division resumed as the temperature increased again. In agreement with cell size and cell number measurements, both microscopic and flow cytometric analyses (see Fig. S3 in the supplemental material) consistently demonstrated an accumu-lation of budded cells in the G2/M phase of circa 70% of the
yeast population at 12°C and subsequent bud release as the
temperature increased (Fig. 6B). Accordingly, analysis of cell cycle phase marker genes (48,49) revealed an enrichment for genes involved in the S/G2phase of the cell cycle (Fig. 6C) at
12°C, while as the temperature increased again, a concerted upregulation of genes involved in early G1and M/G1cell cycle
phases was observed (Fig. 6C).
FIG 5 Discrimination of DTC-specific genes from glucose-responsive genes. (A) Genes in clusters 1, 2, and 3 were pooled (498 genes) and compared to sets of genes identified previously by Kresnowati et al. (589 genes) and Van den Brink et al. (607 genes) to be upregulated upon glucose addition to glucose-limited cultures. P values indicate enrichment for genes identified as being glucose responsive (gray area in the Venn diagram). (B) Genes in clusters 4, 5, and 6 were pooled (604 genes) and compared to sets of genes identified previously by Kresnowati et al. (565 genes) and Van den Brink et al. (1,316 genes) to be downregulated upon glucose addition to glucose-limited cultures. P values indicate enrichment for genes identified as being glucose responsive (gray area in the Venn diagram). (C) For each cluster, the number of genes that were considered to be glucose responsive is indicated in italics, and the number of genes per cluster that showed significantly changed expression in this study only (410 genes in total) is indicated in boldface type. (D) Functional enrichment analysis of the 410 genes excluded from the glucose-responsive genes. MIPS categories are indicated in lightface type, the Gene Ontology category is recognizable by the GO identification number, and enrichment for targets of transcription factors is indicated in boldface type. P values indicate the probability of finding an enrichment by chance.
Hebly et al.
on February 17, 2015 by BIBLIOTHEEK TU DELFT
http://aem.asm.org/
Storage carbohydrate accumulation dynamics during DTC. Metabolism of the storage carbohydrates glycogen and trehalose in S. cerevisiae is known to be affected by temperature (22,59–61). During DTC, the intracellular concentrations of both glycogen and trehalose oscillated. After the temperature maximum, the concentrations of glycogen and trehalose initially decreased (Fig. 7A). Six hours after the temperature maximum, the profiles of these two carbohydrates then diverged. The glycogen concentra-tion increased again after 6 h, while the temperature still de-creased, and kept on increasing until the temperature maximum was reached again. Conversely, the trehalose concentration con-tinued to decrease as the temperature decreased and increased back to its original level only during the 6 h before the temperature maximum. During one DTC, approximately 60% of the net gly-cogen and trehalose content was mobilized and resynthesized.
Levels of UDP-glucose and trehalose-6-phosphate, direct precursors of glycogen and trehalose, respectively, were hardly altered during DTC (Fig. 7B). On the other hand,
glucose-1-phosphate and glucose-6-glucose-1-phosphate, precursors of UDP-glu-cose, displayed an inverse correlation with the applied temper-ature profile (Fig. 7B). Accumulation of reserve carbohydrates could therefore not be correlated directly to intracellular me-tabolite levels in their synthetic pathways. Expression of most genes involved in both synthesis and degradation of glycogen and trehalose during DTC displayed similar expression profiles (Fig. 7C). Typically, their expression levels decreased as the temperature decreased, and the profiles seemed to follow the residual glucose concentration, with a sharp increase in the expression level when the glucose concentration dramatically dropped by⬎80% between 15 and 18 h after the temperature maximum.
Differential physiological and transcriptome responses of S.
cerevisiae during DTC and acclimation to low temperature. To investigate to what extent S. cerevisiae is able to fully acclimatize its physiology and transcriptome to the DTC dynamics imposed upon the chemostat cultures, S. cerevisiae was grown with the TABLE 2 Functional enrichment analysis of DTC-specific genesa
Cluster
Enriched functional category(ies) (MIPS functional category or GO
category) Genes
No. of significantly changed genes in enriched functional category/ total no. of genes in functional
category P value
1 None
2 ER-to-Golgi transport (20.09.07.03)
ERP2, ERP1, YIP3, RER1, SHR3, CHS7
6/72 8.1⫻ 10⫺5
Membrane lipid metabolism (01.06.02)
PLB2, DPM1, CDS1, CST26, CHO1, OPI1
6/83 1.8⫻ 10⫺4
RNA polymerase III transcriptional preinitiation complex assembly (GO:0070898)
NHP6A, NHP6B, YBR090C 3/13 2.7⫻ 10⫺4
Swi4 MNN5, CIS3, PCL1, SVS1, HTA1/
HTA2, PCL2, YHP1, GIC1
8/144 9.4⫻ 10⫺5 Swi6 MNN5, CIS3, PCL1, NRM1, SVS1, PCL2, YHP1, GIC1 8/160 2.0⫻ 10⫺4 Mbp1 SEN34, MNN5, SEC14, PCL1, NRM1, HTA1/HTA2, YHP1, GIC1 8/165 2.4⫻ 10⫺4
Ino2 FAS2, CDS1, KNH1, OPI1 4/31 2.5⫻ 10⫺4
Ino4 YIP3, FAS2, CDS1, CHO1 4/33 3.2⫻ 10⫺4
Tec1 MNN5, PCL1, SVS1, PCL2, CHS7 5/64 4.5⫻ 10⫺4 3 None 4 Degradation of glycine (01.01.09.01.02) GCV2, GCV1 2/5 2.7⫻ 10⫺4 C-1 compound catabolism (01.05.05.07) GCV2, GCV1 2/6 4.1⫻ 10⫺4 Bas1 MTD1, ADE17, GCV2, GCV1 4/39 4.8⫻ 10⫺5
Gcn4 SNO1, GCV2, HIS3, ICY2, YMC1,
GNP1
6/182 3.4⫻ 10⫺4
5 Metabolism of arginine (01.01.03.05)
ARG7, ARG1, CPA1 3/20 4.4⫻ 10⫺4
Arg81 ARG1, CPA1, CUP9 3/22 5.9⫻ 10⫺4
6 Opi1 DAK1, YOP1 2/23 1.7⫻ 10⫺4
a
Each cluster was tested for overrepresentation of genes belonging to various functional categories. MIPS categories are indicated in lightface type, Gene Ontology categories are recognizable by the GO identification number, and enrichment for targets of transcription factors is indicated in boldface type. P values indicate the probability of finding an enrichment by chance.
on February 17, 2015 by BIBLIOTHEEK TU DELFT
http://aem.asm.org/
same experimental setup as was used for the DTC experiments (glucose-limited anaerobic chemostats, with a dilution rate of 0.03 h⫺1) but at a constant temperature of either 30°C or 12°C. The values for physiological characteristics (biomass yield, specific uptake and production rates, and residual glucose concentration) at 30°C and 12°C during DTC closely resembled the steady-state values, with the exception of glycogen contents (Table 3). While trehalose levels were consistently lower at 12°C for both DTC and steady-state cultures, at low temperatures in steady-state cultures, the level of glycogen accumulation was higher than that during DTC (Table 3).
The cell size and budding index (BI) responded similarly in the two systems, with a slight increase in cell size and a strong increase in the BI at 12°C. The BI at 30°C during DTC indicated that 28%⫾ 1% of the population was in the G2/M phase, in comparison with
30%⫾ 3% in steady-state chemostat cultures grown at 30°C. At 12°C during DTC, 72%⫾ 6% of the yeast population was in the G2/M phase, which was close to the BI of 65%⫾ 2% measured at
the same temperature in steady-state chemostat cultures. Principal component analysis (PCA) of DTC and steady-state
transcriptome data showed that transcript levels at 30°C during DTC closely resembled the 30°C steady-state expression levels (Fig. 8). Only 94 genes were differentially expressed between steady-state and DTC cultures at 30°C. Gene expression levels at steady state at 12°C could, however, clearly be distinguished from the levels at 12°C during DTC by PCA (Fig. 8). Pairwise compar-ison of transcript levels at 12°C and 30°C during DTC and in steady-state cultures showed that the response to temperature during DTC (1,061 genes) involved twice as many genes as the temperature response to acclimation in steady-state cultures (521 genes [same P value threshold of 0.05 for both steady-state and DTC data analyses]) (Fig. 9A). However, 44% of the genes that were upregulated and 48% of the genes that were downregulated in steady-state cultures at low temperature responded similarly during DTC (99 upregulated genes and 145 downregulated genes) (Fig. 9B). Genes involved in ribosome biogenesis and targets of Rme1 (indirectly involved in cell cycle control via activation of Cln2 [62]) were upregulated at low temperature, while genes in-volved in the stress response (and particularly targets of Hsf1), protein folding, stabilization, and amino acid transport were FIG 6 Cyclic variation in cell cycle distribution and cell cycle-regulated genes during anaerobic cultivation of S. cerevisiae under diurnal temperature cycles. (A) Cell size (Œ) and cell concentration (). (B) Budding index () and percentage of cells in the G2/M phase of the cell cycle based on flow cytometric analysis (䊐).
The bar illustrates the cell cycle distribution of the population, with light gray indicating the majority population in late G1phase, medium gray indicating the
majority population in S/G2/M phase, and dark gray indicating the majority population in early G1phase. (C) Identification of cell cycle marker genes based on
publically available data sets (46,48,49). a/b* indicates the number of significantly changed genes in the enriched custom-made category/total number of genes in the custom-made category. P values indicate the probability of finding an enrichment by chance. (D) BI () after switching a culture subjected to diurnal temperature cycles back to a fixed temperature of 30°C (black dashed line). Time zero corresponds to the switch. Gray symbols illustrate the BI under diurnal temperature cycles. Error bars represent standard errors of the means of independent duplicate cultures. In all panels, the dashed gray line illustrates the tem-perature profile.
Hebly et al.
on February 17, 2015 by BIBLIOTHEEK TU DELFT
http://aem.asm.org/
downregulated at low temperature. Genes related to phospholipid biosynthesis and the cell cycle also showed a similar response to temperature in DTC and steady-state cultures (see Data Set S4 in the supplemental material). For most of these genes, the magni-tudes of the temperature response were very similar during DTC and steady-state cultures (Fig. 9C). A few genes (6 genes whose expressions were upregulated and 14 genes whose expressions were downregulated at low temperature) did not reach the tran-script level corresponding to low-temperature acclimation during DTC (Fig. 9C). Many genes (393 upregulated and 424 downregu-lated) were specifically responding to the dynamic temperature profiles and not to acclimation. The proteins encoded by those genes were involved mostly in rRNA processing and energy me-tabolism, categories that were associated with the variation in residual glucose concentrations during DTC. Of the 277 temper-ature-responsive genes that were specifically identified in steady-state chemostat cultures, many were involved in transcription ac-tivation, transport of metal ions, regulation of nucleotide metabolism, and the cell cycle (mostly targets of Swi5, a protein that activates transcription of genes expressed at the M/G1phase
boundary and in G1phase [63]) (see Data Set S4 in the
supple-mental material).
DISCUSSION
Diurnal temperature cycling and growth kinetics. Pronounced, rhythmic variations of the residual glucose concentration were among the most prominent phenomena observed during physio-logical analysis of anaerobic, glucose-limited chemostat cultures of S. cerevisiae subjected to DTC. In steady-state glucose-limited chemostat cultures, the residual glucose concentration (CS) is
de-fined by the dilution rate (D) (0.03 h⫺1in this work, which in steady-state chemostats equals the specific growth rate,), the maximum specific growth rate under the experimental conditions (max), and the microorganism’s substrate saturation constant for
glucose (KS). An empirical relation between these variables was
first proposed by Monod (64). ⫽ max·
CS
KS⫹ CS
(3) While did not substantially change during DTC, experimentally determinedmaxvalues at 12°C and 30°C differed by almost 1
order of magnitude (0.310⫾ 0.002 h⫺1and 0.048⫾ 0.001 h⫺1, respectively). Assuming that KSis not affected by temperature
(32), the Monod equation predicts residual glucose concentra-tions, at a specific growth rate of 0.03 h⫺1, of 1.8 mM and 0.1 mM at 12°C and 30°C, respectively. These values closely correspond to the residual concentrations measured at 12°C and 30°C in the DTC experiments (Fig. 2). These observations can be generalized: assuming Monod kinetics and temperature insensitivity of KS, in
any nutrient-limited microbial culture subjected to (diurnal) temperature cycles, irrespective of the identity of the
microorgan-FIG 7 Storage carbohydrate metabolism in S. cerevisiae during diurnal tem-perature cycles. (A) Intracellular glycogen (green circles) and trehalose (blue squares) concentrations during DTC. (B) Intracellular metabolite concentra-tions of key metabolites in storage carbohydrate metabolism, including glucose-1-phosphate (G1P) (orange triangles), glucose-6-phosphate (G6P) (purple squares), UDP-glucose (UDP-gluc) (red circles), and
trehalose-6-phosphate (T6P) (green diamonds). Dotted lines and open symbols in panels A and B represent extrapolated data. (C) Expression profiles of genes coding for enzymes involved in storage carbohydrate metabolism. (D) Overlay of variables relevant for the interpretation of the reserve carbohydrate dynamics, including the residual glucose concentration (blue triangles), qS(dotted red line), budding index (filled circles), temperature (solid red line), and cell cycle distribution (color bar) (Fig. 6). In all panels, error bars represent the standard errors of the means of at least two independent duplicate cultures.
on February 17, 2015 by BIBLIOTHEEK TU DELFT
http://aem.asm.org/
ism or the growth-limiting substrate, CSshould be expected to
fluctuate. Indeed, in a previous study on DTC in glucose-limited cultures, CSwas identified as a key parameter in determining the
dynamics of yeast glycolysis (32). These fluctuations complicate data interpretation of DTC experiments, as nutrient-independent biological responses to DTC may be obscured by cellular re-sponses to changes in CS. Nutrient-independent responses may be
identified by simultaneously studying DTC in cultures with dif-ferent growth-limiting nutrients and subsequently looking for common responses. This approach has been successfully applied to study transcriptional responses to macronutrient limitation and oxygen availability in steady-state chemostat cultures (31,42, 65); however, its application to DTC studies would require many time-consuming and expensive chemostat experiments. As an al-ternative approach, we sought to remove glucose-specific re-sponses from the DTC transcriptome data by comparison with previously reported data sets (50, 51) that capture the effect of glucose on the S. cerevisiae transcriptome. Although those previ-ous studies assessed the impact of glucose in different experimen-tal contexts (aeration, length of exposure to glucose, and/or range of glucose concentrations), circa half of the genes that showed a response to glucose in both previous studies showed a similar response in the DTC cultures (Fig. 5AandB). This correspon-dence demonstrated that the impact of fluctuating glucose con-centrations extends beyond glycolysis and has a major impact on the yeast transcriptome in DTC chemostats.
Transcriptional responses to DTC: beyond glucose. Removal of previously described glucose-responsive genes from the DTC-responsive set of transcripts facilitated interpretation of the scriptome data. However, it may have eliminated genes that tran-scriptionally respond to DTC in glucose-dependent as well as glucose-independent manners. Consistent with temperature-de-pendent remodeling of membrane composition (16,66,67), tran-script levels of genes involved in fatty acid metabolism and, in particular, phospholipid synthesis positively correlated with tem-perature during DTC (Table 2). Cyclic variations of extracellular
TABLE 3 Physiological characteristics of S. cerevisiae grown in glucose-limited anaerobic chemostats a Experimental condition Temp (°C) Ysx (g glucose ·g [dry weight] ⫺ 1) qS (mmol ·g [dry weight] ⫺ 1·h ⫺ 1) qEtOH (mmol ·g [dry weight] ⫺ 1·h ⫺ 1) qCO2 (mmol ·g [dry weight] ⫺ 1 ·h ⫺ 1) Carbon recovery (%) Residual glucose concn (mM)
Glycogen concn (mg glucose equivalen t ·g [dry weight] ⫺ 1) Trehalose concn (mg glucose equivalen t ·g [dry weight] ⫺ 1) Cell size ( m) BI (%) SS 30 0.08 ⫾ 0.004 ⫺ 2.1 ⫾ 0.15 3.2 ⫾ 0.22 3.7 ⫾ 0.10 95 ⫾ 2.2 0.2 ⫾ 0.03 38 ⫾ 0.2 29 ⫾ 0.2 3.6 ⫾ 0.11 30 ⫾ 3.3 12 0.09 ⫾ 0.001 ⫺ 1.8 ⫾ 0.01 2.8 ⫾ 0.01 3.4 ⫾ 0.02 101 ⫾ 0.4 2.1 ⫾ 0.04 121.1 ⫾ 5.7 2.8 ⫾ 0.3 4.4 ⫾ 0.14 65 ⫾ 1.8 DTC 30 0.08 b ⫺ 2.13 c ND 3.9 ⫾ 0.12 ND 0.2 ⫾ 0.01 65 ⫾ 0.7 14.5 ⫾ 0.4 4.0 ⫾ 0.04 28 ⫾ 0.2 12 0.09 b ⫺ 1.96 c ND 3.0 ⫾ 0.11 ND 2.6 ⫾ 0.07 50.5 ⫾ 3.6 7.7 ⫾ 0.6 4.4 ⫾ 0.05 72 ⫾ 6.4 aValues represent the averages ⫾ standard errors of the mean of at least two independent replicates. SS, steady state; EtOH, ethyl alcohol; ND, not determined. bThe biomass yield during DTC was calculated by using the biomass specific glucose consumption rate listed and a specific growth rate of 0.03 h ⫺ 1. cThe profile of the biomass specific glucose consumption rate during DTC is shown in Fig. 3A .The intermediate qS values of the time intervals of ⫺ 1.5 h to 1.5 h and 10.5 to 13.5 h, corresponding to the qS at 30°C and 12°C, respectively, are shown.
FIG 8 Principal component analysis of DTC and steady-state transcriptome data. The plot shows the average expression levels of all genes at each sampling point projected onto the first and second principal components (PC1 and PC2). , sampling points during DTC; , 12°C steady-state (SS) sampling point; Œ, 30°C steady-state sampling point.
Hebly et al.
on February 17, 2015 by BIBLIOTHEEK TU DELFT
http://aem.asm.org/
FIG 9 Comparative transcriptomics of the temperature responses during acclimation and DTC. (A) Two pairwise comparisons between transcript levels at 30°C and those at 12°C were performed. Shown are data for one comparison of transcript levels at 30°C at steady state (orange filled circle) versus 12°C at steady state (open light blue circle) and one comparison of 30°C during DTC (red filled circle) versus 12°C during DTC (open dark blue circle). The gray dots represent the sampling points for microarray analysis. (B) The sets of temperature-responsive genes from acclimated and DTC cultures were subsequently compared after pooling of the genes according to their temperature response (left, upregulated at 12°C; right, downregulated at 12°C). (C) Genes that displayed similar temperature responses in both acclimated and DTC cultures were clustered according to the magnitude of their changes in expression levels, and box plots illustrate the mean-normalized expression level per cluster. The change in expression levels of 83 genes was more pronounced during DTC than in acclimated cultures (top), the majority of the genes (141 genes) responded to temperature during DTC and acclimation with similar magnitudes (middle), and 20 genes did not reach full acclimation expression levels at 12°C during DTC (bottom).
on February 17, 2015 by BIBLIOTHEEK TU DELFT
http://aem.asm.org/
acetate concentrations (Fig. 3H) may be related to the role of acetyl coenzyme A (acetyl-CoA) as a key precursor for fatty acid synthesis (16,68). Genes involved in ER-to-Golgi transport were upregulated as the temperature decreased during DTC (Table 2). Since the yeast plasma membrane and cell wall composition are both affected by temperature (66,69), this response may reflect the need for continuous replacement of membrane and cell wall components. Intracellular membrane trafficking is also sensitive to temperature in mammalian cells (70–72), and in a genome-wide screen for sensitivity to high pressure and low temperature, yeast mutants affected in membrane trafficking showed decreased fitness (73). A possible temperature-dependent alteration of the cell wall during DTC is further supported by the overrepresenta-tion of target genes of the transcripoverrepresenta-tion factors Swi4 and Swi6, which are known to play an important role in the regulation of genes required for cell wall remodeling (69), among the DTC-responsive transcripts.
Genes involved in arginine synthesis (ARG7, ARG1, and CPA1) and glycine degradation (GCV2 and GCV1) showed decreased transcript levels when temperature was decreased during DTC. Regulated by the transcription factors Arg81 and Bas1, respec-tively, both these processes are also controlled by Gcn4 (74). How-ever, despite a positive correlation of GCN4 expression with tem-perature (18, 75), only a few of its target genes (74) displayed temperature-dependent expression. Genes involved in de novo purine synthesis, a process regulated by Bas1 and Gcn4 (74), were also overrepresented among genes whose expression levels were reduced at 12°C. An upregulation of these genes upon a relief of S.
cerevisiae cultures from glucose limitation was previously
corre-lated with a strong decrease in intracellular adenine nucleotide concentrations (50). However, the adenine nucleotide concentra-tion remained constant during the DTC (see Fig. S4 in the supple-mental material). The fluctuations of the expression of genes in-volved in de novo purine biosynthesis may therefore reveal a regulation specific to dynamic temperature conditions.
Cell cycle synchronization during DTC: temperature or rel-ative specific growth rate? Rhythmic variations of the budding index, flow cytometric analysis, and transcription of cell cycle-responsive genes all indicated a partial synchronization of the yeast cell cycle during the DTC experiments. In this respect, it is relevant to note that at the dilution rate used for chemostat culti-vation (D⫽ 0.03 h⫺1), the average biomass doubling time (td⫽
ln2/ ⫽ ln2/D) of 23.1 h closely corresponded to the length of the 24-h temperature cycle, which might have facilitated partial syn-chronization. Rapid loss of synchronization was observed upon release from DTC (Fig. 6D). This observation showed that under the experimental conditions, DTC caused an externally imposed cell cycle synchronization rather than “entrainment” of an endog-enous synchronization mechanism (e.g., by a yeast circadian clock [76], for which there is no molecular evidence at present). Loss of synchronization upon release from DTC in anaerobic chemostat cultures also represents a marked difference from the cell cycle-related spontaneous and sustained metabolic oscillations that are frequently observed in aerobic, glucose-limited chemostat cul-tures of S. cerevisiae (77–80).
A much higher incidence of G2/M cells at 12°C than at 30°C
was observed not only during DTC but also in nonsynchronized, steady-state chemostat cultures grown at the same dilution rate (Table 3). A possible interpretation of these observations is that at low temperatures, progression through the G2/M phase is slowed
down more strongly than other phases of the cell cycle. However, from a previous study on batch cultures of S. cerevisiae, it was de-duced that temperature had only a minor impact on the budding index (81). Conversely, chemostat studies performed at a constant temperature of 30°C revealed a strong impact of the specific growth rate on the budding index and cell cycle distribution (43, 82,83). These seemingly contradictory observations can be recon-ciled if cell cycle distribution in S. cerevisiae is primarily correlated not with temperature or the absolute specific growth rate but in-stead with the relative specific growth rate (/max) under a given
set of cultivation conditions. This would imply that in batch cul-tures ( ⬇ max) grown at different temperatures, no impact on
cell cycle distribution is expected, as observed previously by Vanoni et al. (81). Conversely, variation of at a constant tem-perature (i.e., at constantmax) (43,82,83) or
temperature-im-posed variation ofmaxat constant (this study) should affect the
cell cycle distribution. Our results with DTC and steady-state che-mostat cultures (D⫽ 0.03 h⫺1), in which the/maxvaried from
approximately 0.1 at 30°C to 0.63 at 12°C, are in full agreement with this hypothesis.
Dynamics of storage carbohydrate content during diurnal temperature cycles. Intracellular pools of glycogen and trehalose, the two major storage carbohydrates in S. cerevisiae, displayed intriguing dynamics during the DTC. In S. cerevisiae, levels of these two glucose polymers are affected by many parameters, including nutrient availability (84) (and thereby growth rate [43, 82]), temperature, and cell cycle progression (reviewed in refer-ences59and85). Interpretation of the dynamics of storage carbo-hydrate pools during DTC is complicated due to the strong vari-ations of all three of these parameters. Cold shock (23) and full acclimation (Table 3) have been shown to affect storage carbohy-drate accumulation. However, no direct correlations of intracel-lular contents of glycogen and trehalose with temperature were observed during DTC, and it is likely that temperature per se was not the primary trigger for the reserve carbohydrate dynamics during DTC.
Under slow-growth, aerobic, glucose-limited conditions, S.
cerevisiae stores glycogen and trehalose in G1phase and quickly
mobilizes them in late G1phase to provide metabolic energy for
budding (77). During DTC, mobilization of reserve carbohy-drates coincided with a phase of the temperature cycle where the majority of the population was in late G1phase (Fig. 7D). This
observation is consistent with the notion that during DTC, reserve carbohydrate dynamics were governed primarily by cell cycle pro-gression. However, this interpretation is likely to be an oversim-plification, since the dynamics of trehalose and glycogen clearly differed, probably reflecting the involvement of different signaling pathways. Resolving the mechanism by which storage carbohy-drate metabolism is regulated during DTC will require detailed analysis of culture heterogeneity, especially with respect to cell cycle progression.
Yeast adaptation to diurnal temperature variation approaches full acclimation. At the temperature extremes (12°C and 30°C) during DTC, the physiology of anaerobic S. cerevisiae chemostat cultures resembled that of steady-state chemostat cultures that were fully acclimatized to these two temperatures (Table 3). At 30°C, transcript profiles of steady-state and DTC cultures could hardly be discriminated by principal component analysis (Fig. 8). While a larger difference was observed at 12°C, low-temperature-responsive genes were involved in similar cellular functions under
Hebly et al.
on February 17, 2015 by BIBLIOTHEEK TU DELFT
http://aem.asm.org/
steady-state and DTC regimes (lipid biosynthesis, cell cycle, stress response, arginine metabolism, and various cellular functions). The response to dynamic temperature fluctuations could there-fore largely be predicted from acclimation experiments. Two cat-egories of genes specifically responded to DTC, i.e., ER-to-Golgi transport and de novo purine biosynthesis, suggesting that the dynamic nature of the temperature profile rather than the temper-ature triggered this response.
Comparison of the DTC-specific transcriptome response to the transcriptome response to instantaneous cold shock revealed that only very few genes showed a similar transcriptional response to these two dynamic temperature profiles. Only 10 genes were downregulated during DTC as well as in three previously reported cold shock studies (22, 23, 61) (MTD1, GCV2, HCH1, AQR1,
DSE4, ICY2, FUI1, GCV1, GNP1, and PCL5), and 8 genes were
upregulated (TOS4, IZH1, IZH4, KNH1, BIO2, PAU10, YLR225C, and YHR138C). The consistent downregulation of genes involved in de novo purine biosynthesis (represented by GCV1 and GCV2) in all these dynamic studies, but not during experiments at a fixed temperature, suggests a role for this cellular function during dy-namic temperature profiles. IZH1 and IZH4, which are both in-volved in intracellular zinc ion homeostasis, were upregulated at low temperature during DTC and cold shock. Interestingly, com-parison of the transcriptome response to cold shock with the tran-scriptome response during acclimation to low temperatures also identified genes involved in zinc ion transport. This suggests a relationship between zinc homeostasis and temperature. This hy-pothesis is supported by a previous report on temperature-depen-dent accumulation of Zn2⫹by S. cerevisiae (77,86) and the well-documented role of zinc in regulation of membrane fluidity (87). In summary, growth kinetics of anaerobic, glucose-limited chemostat cultures of S. cerevisiae subjected to DTC could be de-scribed by simple Monod kinetics. Rhythmic changes of the resid-ual glucose concentration, which are inevitable consequences of these kinetics, had a strong impact on the yeast transcriptome during DTC. The physiology and transcriptome during DTC re-sembled those of steady-state, temperature-acclimated chemostat cultures but showed a minor overlap in yeast cells exposed to instantaneous (“cold shock”) temperature changes. A partial, ex-ternally imposed synchronization of the yeast cell cycle in DTC chemostat cultures was not caused primarily by fluctuations of the temperature. Instead, comparison with data reported in the liter-ature suggested that fluctuations in the relative specific growth rate (/max) caused by DTC underlie the observed
synchroniza-tion. In the present study, the specific growth rate of the cultures was kept virtually constant. In natural environments, the specific growth rate can vary during DTC. To capture such growth dy-namics, future studies may be based on alternative (semi)continu-ous cultivation setups such as repeated batch cultures, fed-batch cultures, or auxostats (32,88).
ACKNOWLEDGMENTS
We give special thanks to Danilo Porro and Frank Rosenzweig and their coworkers for their help with the flow cytometric analysis. We thank Marinka Almering, Markus Bisschops, Sebastiaan de Bruin, Xavier Hak-kaart, Marijke Luttik, Mathijs Martens, and Tim Vos for technical assis-tance.
This project was carried out within the research programs of the Kluyver Centre for Genomics of Industrial Fermentation and the
Nether-lands Consortium for Systems Biology, which are both sponsored by the Netherlands Genomics Initiative.
REFERENCES
1. Madigan MT, Martinko JM. 2006. Brock biology of microorganisms. Prentice Hall, Upper Saddle River, NJ.
2. Feller G, Gerday C. 2003. Psychrophilic enzymes: hot topics in cold adaptation. Nat. Rev. Microbiol. 1:200 –208.http://dx.doi.org/10.1038 /nrmicro773.
3. Thieringer HA, Jones PG, Inouye M. 1998. Cold shock and adaptation. Bioessays 20:49 –57.
4. Mansilla MC, Cybulski LE, Albanesi D, de Mendoza D. 2004. Control of membrane lipid fluidity by molecular thermosensors. J. Bacteriol. 186: 6681– 6688.http://dx.doi.org/10.1128/JB.186.20.6681-6688.2004. 5. Schumann W. 2009. Temperature sensors of eubacteria. Adv. Appl.
Mi-crobiol. 67:213–256.http://dx.doi.org/10.1016/S0065-2164(08)01007-1. 6. Gasch AP, Werner-Washburne M. 2002. The genomics of yeast
re-sponses to environmental stress and starvation. Funct. Integr. Genomics 2:181–192.http://dx.doi.org/10.1007/s10142-002-0058-2.
7. Dragosits M, Stadlmann J, Albiol J, Baumann K, Maurer M, Gasser B, Sauer M, Altmann F, Ferrer P, Mattanovich D. 2009. The effect of temperature on the proteome of recombinant Pichia pastoris. J. Proteome Res. 8:1380 –1392.http://dx.doi.org/10.1021/pr8007623.
8. Gunnigle E, McCay P, Fuszard M, Botting CH, Abram F, O’Flaherty V. 2013. A functional approach to uncover the low-temperature adaptation strategies of the archaeon Methanosarcina barkeri. Appl. Environ. Micro-biol. 79:4210 – 4219.http://dx.doi.org/10.1128/AEM.03787-12. 9. Marr AG, Ingraham JL. 1962. Effect of temperature on the composition
of fatty acids in Escherichia coli. J. Bacteriol. 84:1260 –1267.
10. Jones PG, Inouye M. 1994. The cold-shock response—a hot topic. Mol. Microbiol. 11:811– 818. http://dx.doi.org/10.1111/j.1365-2958 .1994.tb00359.x.
11. Al-Fageeh MB, Smales CM. 2006. Control and regulation of the cellular responses to cold shock: the responses in yeast and mammalian systems. Biochem. J. 397:247–259.http://dx.doi.org/10.1042/BJ20060166. 12. Lindquist S. 1986. The heat-shock response. Annu. Rev. Biochem. 55:
1151–1191.http://dx.doi.org/10.1146/annurev.bi.55.070186.005443. 13. Klinkert B, Narberhaus F. 2009. Microbial thermosensors. Cell. Mol. Life
Sci. 66:2661–2676.http://dx.doi.org/10.1007/s00018-009-0041-3. 14. Postmus J, Canelas AB, Bouwman J, Bakker BM, van Gulik WM, de
Mattos MJT, Brul S, Smits GJ. 2008. Quantitative analysis of the high temperature-induced glycolytic flux increase in Saccharomyces cerevisiae reveals dominant metabolic regulation. J. Biol. Chem. 283:23524 –23532.
http://dx.doi.org/10.1074/jbc.M802908200.
15. Salvado Z, Arroyo-Lopez FN, Guillamon JM, Salazar G, Querol A, Barrio E. 2011. Temperature adaptation markedly determines evolution within the genus Saccharomyces. Appl. Environ. Microbiol. 77:2292–2302.
http://dx.doi.org/10.1128/AEM.01861-10.
16. Redon M, Guillamon JM, Mas A, Rozes N. 2011. Effect of growth temperature on yeast lipid composition and alcoholic fermentation at low temperature. Eur. Food Res. Technol. 232:517–527.http://dx.doi.org/10 .1007/s00217-010-1415-3.
17. Lodolo EJ, Kock JL, Axcell BC, Brooks M. 2008. The yeast Saccharomyces cerevisiae—the main character in beer brewing. FEMS Yeast Res. 8:1018 – 1036.http://dx.doi.org/10.1111/j.1567-1364.2008.00433.x.
18. Pizarro FJ, Jewett MC, Nielsen J, Agosin E. 2008. Growth temperature exerts differential physiological and transcriptional responses in labora-tory and wine strains of Saccharomyces cerevisiae. Appl. Environ. Micro-biol. 74:6358 – 6368.http://dx.doi.org/10.1128/AEM.00602-08. 19. Sasano Y, Haitani Y, Hashida K, Oshiro S, Shima J, Takagi H. 2013.
Improvement of fermentation ability under baking-associated stress con-ditions by altering the POG1 gene expression in baker’s yeast. Int. J. Food Microbiol. 165:241–245.http://dx.doi.org/10.1016/j.ijfoodmicro.2013.05 .015.
20. Gonçalves P, Valerio E, Correia C, de Almeida JM, Sampaio JP. 2011. Evidence for divergent evolution of growth temperature preference in sympatric Saccharomyces species. PLoS One 6:e20739.http://dx.doi.org /10.1371/journal.pone.0020739.
21. Hittinger CT. 2013. Saccharomyces diversity and evolution: a budding model genus. Trends Genet. 29:309 –317.http://dx.doi.org/10.1016/j.tig .2013.01.002.
22. Murata Y, Homma T, Kitagawa E, Momose Y, Sato MS, Odani M,
